自组织神经网络地震岩相分析

Seismic Lithofacies Analysis with Self-Organizing Neural Network

自组织神经网络据研究对象属性特征的相似性,将对象按要求类别数对分类进行正确归类,避免了BP神经网络必须提供学习样本的缺点。通过地震数据自动识别地震岩相,在塔里木盆地LGX油田应用取得了明显效果。在岩相图上解释了6条近东西向展布的断裂,这些断裂控制了LGX油田的油气分布,与实钻结果相符,说明自组织神经网络碳酸盐岩储集层的有效性。

Self-organizing neural network (SONN) could be used to properly classify the studied objects according to the similarity of attributes from the objects, and avoid the shortcoming of back propagation neural network (BPNN) that learning-samples have to be provided. The method by using SONN to auto-identify the seismic lithofacies from seismic data has been successfully applied in LGX oilfield of Tarim basin. From the lithofacies map, 6 faults controlled the hydrocarbon distribution in LGX oilfield, with extending in near east-west orientation, are revealed by using SONN method, which accord with the results from drilled wells. The study shows the feasibility of SONN in carbonate reservoirs.